Overexpression analysis

heat_with_all_greater %>% melt(.) %>%
  mutate(gene = rep(c("ALK", "ARHGAP26", "ERG", "ETV1", "MAML3", "NTRK3", "RARA", "RET", "TACC3", "TFE3"), times = 19), .before = variable) %>%
  mutate(groups = cut(.$value, breaks = c(1, 0.05, 0))) %>%
  ggplot(., aes(gene, variable, fill = groups)) +
  geom_tile( colour = "white", linetype = 1) +
  scale_fill_discrete(breaks = levels(groups)) +
  theme(axis.text = element_text(size = 15)) +
  theme(axis.text.x = element_text(angle = 45, vjust = 0.5, hjust=0.5)) +
  labs(x = "Gene", y = "Cancer Type") 

No id variables; using all as measure variables

comparison of overexpression approaches

95th percentile overexpression identified more fusion gene samples only 3 samples were identified by Tukey's outlier that were not identified by 95th percentile (3 ETV1 prostate adenocarcinoma samples)

false positives when applied to all cancers

outlier_false_pos <- all_genes %>%
  filter(is_outlier == TRUE, fusion != "Fusion") %>% select(rail_id, gene, Project_ID)
write_csv(outlier_false_pos, 'outlier_false_pos.csv')

above95_false_pos <- all_genes %>%
  filter(above_95 == TRUE, fusion != "Fusion") %>% select(rail_id, gene, Project_ID)
write_csv(above95_false_pos, 'above95_false_pos.csv')

shared_false_pos <- all_genes %>%
  filter(above_95 == TRUE, is_outlier == TRUE, fusion != "Fusion") %>% select(rail_id, gene, Project_ID)
write_csv(shared_false_pos, 'shared_false_pos.csv')

all_genes %>%
  ungroup() %>%
  filter(has_cancer == "Cancer", above_95 == TRUE, fusion != "Fusion") %>% select(rail_id, gene, Project_ID) %>%
  group_by(gene) %>%
  mutate(gene_total = n()) %>%
  distinct(gene_total)

ided_by_both 
outlier_fusions %>% filter(gene != "TMPRSS2")
above_95percent

all_genes %>% filter(fusion == "Fusion") %>% ungroup() %>%
  summarise(n = n())

ided_by_both %>% ungroup() %>% group_by(gene) %>% summarise(n = n())

outlier_fusions %>% ungroup() %>% group_by(gene) %>% summarise(n = n())

difference is outlier picks up an extra 3 in ETV1

ided_by_both %>% filter(gene == "ETV1")

outlier_fusions %>% filter(gene == "ETV1")

-it picks up 3 extra etv1 prostate fusions

etv1_combined %>% filter(has_cancer == "Cancer", Project_ID == "PRAD") %>%
  ggplot(aes(x = Project_ID, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = is_outlier))

etv1_combined %>% filter(has_cancer == "Cancer", Project_ID == "PRAD") %>%
  ggplot(aes(x = log2_auc_norm)) +
  geom_density() +
  geom_vline(aes(xintercept = log2_auc_norm), data = . %>% filter(fusion == "Fusion"))

shape_in_words3(shape_for_heatmap(etv1_combined))

outlier_false_pos
above95_false_pos
shared_false_pos

-larger discrepancy in false pos

comparing false positives

fpr_specific %>% group_by(gene, Project_ID) %>% summarise(n = n())

1241 samples for 95th

fpr_specific_outlier %>% group_by(gene, Project_ID) %>% summarise(n = n())

322 for outlier

fpr_specific_shared %>% group_by(gene, Project_ID) %>% summarise(n = n())

282 shared

overexp captures capture all the same for alk, arhgap26, erg, etv1 prad shared 19 (outlier 33), maml3, ntrk3, rara, ret luad shared 28 (outlier 54), tacc3, tfe3

so 2 cases where overexpress doesn't capture same false positives

etv1_combined %>% filter(has_cancer == "Cancer", Tissue == "Prostate", fusion == "Fusion")

boxplot visualisation of genes

alk_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

arhgap26_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

erg_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

etv1_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

maml3_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

ntrk3_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

rara_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

ret_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

tacc3_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

tfe3_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

tmprss2_combined %>%
  filter(cancer_status == "Cancer") %>%
  ggplot(aes(x = Cancer, y = log2_auc_norm)) +
  geom_boxplot() +
  geom_jitter(aes(colour = Cancer), data=. %>% filter(fusion == "Fusion")) +
  theme(legend.position = "none") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

Tukey's outlier method

so these graphs are the the outlier == TRUE

expression_function_mutate(erg_counts) %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(maml3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(tfe3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(tmprss2_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(rara_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

#middle
expression_function_mutate(tacc3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(ret_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(etv1_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(ntrk3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(alk_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(arhgap26_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(is_outlier == "TRUE") %>%
  ggplot(aes(x = Cancer, fill = fusion)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

graphs fusions == TRUE, stacked by outliers

#early breaks
expression_function_mutate(erg_counts) %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(maml3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(tfe3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(tmprss2_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(rara_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

#middle
expression_function_mutate(tacc3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(ret_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(etv1_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(ntrk3_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(alk_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

expression_function_mutate(arhgap26_counts) %>%
  filter(cancer_status == "Cancer") %>%
  filter(fusion == "Fusion") %>%
  ggplot(aes(x = Cancer, fill = is_outlier)) +
  geom_bar(position = "fill") +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

-not as nice looking like this